Comp541 - Machine Learning Term Project
Orijinal Paper:
Ronghang Hu, Marcus Rohrbach, Jacob Andreas, Trevor Darrell, Kate Saenko, 2016, Modeling Relationships in Referential Expressions with Compositional Modular Networks, Web, url: https://arxiv.org/abs/1611.09978
Aim:
Replication of the above paper as a machine learning class ter project
Summary:
Localization of referential expressions in an image
Abstract:
Generally, when we are talking about some people or some object, we refer them accordingly their relationship with the other objects or people. However, if we, as a human, understand objects or people with these relationships, then computers also must be able to understand this way so that we can have more accurate results. In this paper, I use Computational Modular Networks (CMNs) which first parses the referential expression as entities and relations,then identifies in the image and finds the most related one.